Scientific Blog

Welcome to my scientific blog, a digital diary where I share my insights, experiences, and discoveries in the ever-evolving world of artificial intelligence. Here, you'll find a collection of articles, discussions, and reflections that encapsulate my journey through the intricacies of AI research and its practical applications.

What You will find here

Featured Topics

Dive into the cutting-edge world of deep learning with my latest tutorial on leveraging the NVIDIA Container Toolkit for Docker. This beginner-friendly guide will walk you through setting up your environment to use GPUs for training deep learning models in Docker containers, making your AI projects faster and more efficient. Whether you're an AI researcher, a data scientist, or a developer eager to harness the power of GPUs, this blog will provide you with the essential steps to get started. 

Foundation Models and Combinatorial Optimization

The ANT system is a theoretical Large Language Model (LLM) uniquely designed for Optimization tasks. It takes inspiration from the recent advancements in specialized open-source LLMs tailored for fields like medicine and math. Essentially, the ANT model is an imagined entity skilled in solving optimization problems. This discussion will focus on the potential benefits of developing such a model and the obstacles that might be encountered in the process.

Artificial Curiosity in Meta-Heuristics

Meta-heuristics have long been the cornerstone of solving complex combinatorial optimization problems. However, as we venture deeper into the realm of artificial intelligence, it becomes evident that traditional methods need a fresh perspective. In this blog post, we explore how the integration of Artificial Curiosity and Active Inference can revolutionize perturbation techniques in classical meta-heuristics, leading to more efficient and effective problem-solving strategies.

Streamlining Space in Routing Problem 

In this blog post, we explore the critical aspect of reducing solution space in routing problems, a key challenge in combinatorial optimization. The post begins by emphasizing the importance of minimizing solution space to improve the efficiency and accuracy of solving routing problems. Then, it compares different methods, such as heuristic and metaheuristic techniques, machine learning applications, and advanced mathematical models, to see how well they work at reducing the size of the solution space.

Machine Learning meets Heuristic Methods

This blog post delves into the integration of Machine Learning with Constructive Heuristics, Local Searches, and Meta-heuristics to effectively address Combinatorial Optimization (CO) problems. It specifically focuses on maintaining efficient scaling and performance for larger problem sets. The author illustrates these concepts through his research at IDSIA, tackling the renowned Traveling Salesman Problem. Additionally, the post discusses how the insights gained from this work can be applied to a broader range of CO challenges.

Join the conversation! I believe in the power of knowledge sharing and community engagement. I encourage you to dive into the discussions, share your thoughts, and be part of a growing community passionate about the future of AI.